It can find and use zero-day bugs on its own, which puts it in OpenAI's top risk bucket. The model will roll out soon but its strongest hacking abilities will be locked down with extra checks.
Join the thread
Discussion
Sign in to join the discussion.